The aim of this study was to investigate the reliability of a protocol designed to simulate endurance performance in events of long duration (approximately 5 h) where endogenous carbohydrate stores are low. Seven male subjects were recruited (age 27 +/- 7 years, VO(2max) 66 +/- 5 ml/kg/min, W (max) 367 +/- 42 W). The subjects underwent three trials to determine the reliability of the protocol. For each trial subjects entered the laboratory in the evening to undergo a glycogen-depleting exercise trial lasting approximately 2.5 h. The subjects returned the following morning in a fasted state to undertake a 1-h steady-state ride at 50% W (max) followed by a time trial of approximately 40-min duration. Each trial was separated by 7-14 days. The trials were analysed for reliability of time to completion of the time trial using a coefficient of variation (CV), with 95% confidence intervals (data are mean +/- SD). The times to complete the three trials were 2,546 +/- 529, 2,585 +/- 490 and 2,568 +/- 555 s for trials 1, 2 and 3, respectively. The CV between trials 1 and 2 was 4.5% (95% CI 2.9-10.4%) and between trials 2 and 3, 3.8% (95% CI 2.4-9.9%). There was no difference in oxygen uptake, respiratory exchange ratio, carbohydrate oxidation, fat oxidation, plasma glucose concentration and plasma lactate concentration between the three trials. Therefore we can conclude that prior glycogen depletion does produce a reliable measure of performance with metabolic characteristics similar to ultraendurance exercise.